Learning OLL without drowning
A concrete order for the fifty-seven cases, why the dots come last, and how to drill a group until it stops needing thought.
By the end you will be able to
- Take the 57 cases in an order that keeps the payoff coming
- Drill a group of two to six cases until recognition and execution are both automatic
- Keep solving at full speed while half the set is still unlearnt
- Say why the eight dot cases are worth the least per algorithm
Fifty-seven cases is not a memory problem, it is a scheduling problem. People who stall do not stall because their memory failed; they stall because they took the cases in an order that made every week feel the same as the last one, with no group ever quite finished.
The order below fixes that. It starts with the groups you can finish in a sitting, keeps every mirror pair together, and leaves the six L shapes and the eight dots until you have months of momentum behind you.
The seven you already have
The corner algorithms from two-look OLL — sune, anti-sune, headlights, T, bowtie, pi and H — are the seven cases where every edge is already oriented. They are OLL 21 to 27 and they need no further work.
Those seven cover one solve in eight. Hold that figure next to another one: the eight dot cases, the longest algorithms in the set, also cover one solve in eight between them. Seven cases you already have against eight you have not started, for exactly the same share of your solving. That comparison decides the order everything else goes in.
The order
| Order | Group | Cases | Why here |
|---|---|---|---|
| — | All edges oriented | 7 (OLL 21–27) | Already known from two-look |
| 1 | Corners oriented | 2 (28, 57) | Two cases, both unmistakable from the top face alone |
| 2 | Squares | 2 (5, 6) | A mirror pair, seven turns each, one shape |
| 3 | T shapes | 2 (33, 45) | You already use 45 as the two-look line algorithm |
| 4 | C shapes | 2 (34, 46) | Two more cases with one shape between them |
| 5 | W shapes | 2 (36, 38) | A mirror pair built from turns you have |
| 6 | P shapes | 4 (31, 32, 43, 44) | Two mirror pairs; 43 and 44 are six turns each |
| 7 | Fish shapes | 4 (9, 10, 35, 37) | Sune-shaped, and the recognition is distinctive |
| 8 | Lightning bolts | 6 (7, 8, 11, 12, 39, 40) | Three mirror pairs, two of them seven turns each |
| 9 | Knight move shapes | 4 (13, 14, 15, 16) | Two mirror pairs that need care told apart |
| 10 | Awkward shapes | 4 (29, 30, 41, 42) | Two mirror pairs, longer algorithms |
| 11 | L shapes | 6 (47, 48, 49, 50, 53, 54) | Six cases, one shape between them |
| 12 | I shapes | 4 (51, 52, 55, 56) | Four cases, one shape, wide turns throughout |
| 13 | Dots | 8 (1, 2, 3, 4, 17, 18, 19, 20) | Rarest per case, longest to execute |
The two that come first
Both cases in the first group have every corner already oriented, which means the algorithm only has to flip edges. They are also two of the seven cases in the whole set that can be told apart from the top face alone, so they cost you nothing in recognition.
M' in the middle is the only awkward part.A mirror pair, done properly
The squares are the first place to practise learning a case and its mirror together rather than months apart. They are the same seven turns with every right becoming a left and every direction reversed.
Drilling a group
This is the part that decides whether the eight months work. The loop below takes about twenty minutes a day for a pair of cases, and about a week for a group of four.
- Work out the fingers before the speed. Run the algorithm at one turn a second and settle which finger makes each turn. Count the regrips; if there is more than one, look at the other versions on the set page before you commit.
- Set the case up from solved. Run the algorithm backwards — every turn reversed, in reverse order — and you are looking at the case. The trainer does this for you, which is the whole reason it exists.
- Thirty slow repetitions of the new case. Same fingers each time. Slow means slower than feels sensible.
- Then the whole group shuffled. This is the step people skip and it is the step that builds recognition. Set up a random case from the group, name it out loud, then solve it. Naming before turning is not optional — that is the skill you are buying.
- Sleep on it. Do the group cold the next morning before anything else. Anything you have to look up was not learnt, and it goes back to step three.
- Only then, into solves. Keep two-look ready as a fallback: if you cannot name the case within about two seconds, do the two-look route and carry on. Stalling mid-solve to dig for an algorithm teaches you to stall.
- Leave the group alone for a fortnight, then test it cold. What survives that is yours.
Why the dots come last
The dot cases — no yellow edge facing up at all — are eight of the fifty-seven, and between them they account for one solve in eight. That is the same share the seven cross cases cover, and the cross cases were free.
They are also individually the rarest things in the set. Most OLL cases turn up about once in fifty-four solves. OLL 1 turns up once in a hundred and eight, and OLL 20 — every corner oriented, every edge flipped — once in two hundred and sixteen, which makes it the rarest case in CFOP. You could learn it in March and not meet it until June.
And they are the longest. Every dot algorithm in the library runs to eleven, twelve or thirteen turns, against six or seven for the squares and the shorter P shapes. Eight long algorithms, learnt last, for the least frequent eighth of your solves.
Keeping the solve honest while you learn
- Do not stop timing. You want to see the dip when a group goes in and the recovery a fortnight later; that curve is what tells you the pace is right.
- Do not learn a new group in the week you are also working on lookahead. One new thing at a time, always.
- If a case has been in your set for a month and you still hesitate, the problem is recognition, not memory. The next lesson is about that.
- Ten solves a day with a clean two-look last layer beat fifty solves a day spent hunting for half-learnt cases.